Research

Open Credit Scoring explores how causal AI, systems thinking, and open standards can help build the next generation of trustworthy financial decision systems.

Our research focuses on moving beyond purely statistical machine learning approaches toward AI systems that can reason about causation, accountability, interventions, and long-term system behavior.

From proprietary black boxes to open scientific infrastructure

Next-generation financial AI cannot be built only through closed model development. It requires shared benchmarks, transparent assumptions, interoperable governance frameworks, and reproducible research. Open Credit Scoring focuses on the scientific and technical infrastructure needed to move credit decision systems beyond opaque prediction.

From Machine Learning to Causal AI

Most modern AI systems in finance are based primarily on statistical prediction.

These systems identify patterns and correlations in historical data, but they often struggle to distinguish between correlation and causation. As a result, they may unintentionally reinforce historical biases, produce unstable outcomes, or create systems that are difficult to govern and explain.

Our research explores how causal inference and causal models can augment machine learning systems by introducing explicit representations of cause-and-effect relationships.

Causal Bayesian networks

Causal fairness graphs

Counterfactual reasoning

Interventions and policy simulation

Causal debiasing techniques

Systems thinking and feedback loops

From Fairness Metrics to Causal Fairness

Many existing AI fairness approaches rely on statistical fairness metrics.

While useful, these metrics can become mathematically incompatible with one another and may fail to distinguish between legitimate causal relationships and discriminatory effects.

Rather than treating fairness as a purely statistical optimization problem, we explore fairness as a causal and institutional problem that must be understood within broader social and economic systems.

Disparate treatment

Disparate impact

Proxy discrimination

Alternative data usage

Digital redlining

Long-term feedback effects in financial systems

From Explainability to Governance

Traditional explainability approaches often focus on interpreting black-box machine learning models after they have already been trained.

Our research explores a different approach: using causal models as operational governance architectures for high-stakes AI systems.

Instead of only asking: “How do we explain the model?”

We also ask: “How do we govern and constrain the system itself?”

Policy constraints

Regulatory compliance

Human oversight

Institutional accountability

Transparent decision processes

Simulation of downstream effects over time

From Prediction to Control and Simulation

Most machine learning systems are optimized primarily for prediction.

Our research investigates how AI systems can evolve beyond prediction toward control of decision systems through causal inference and simulation of long-term system behavior through systems thinking and system dynamics.

This expands the role of AI from passive prediction toward active governance and policy design for high-stakes domains such as credit underwriting and financial services.

Research Areas

Current research areas include:

Causal AI for credit scoring and underwriting

AI fairness and antidiscrimination law

Alternative data and causal debiasing

Causal explainability and transparency

Systems thinking for AI governance

Open technical standards for trustworthy AI

High-stakes decision-making systems

Human-centered and institutionally accountable AI

Research Vision

We believe the future of financial AI requires more than increasingly complex black-box models.

It requires systems that are scientifically grounded, causally informed, transparent, and governable by design.

Open Credit Scoring exists to help advance that transition.